A method and system for predicting the harvest period of *Hymenochloa crus-galli* by integrating multi-source data

Abstract

This invention relates to the field of data reasoning technology, and discloses a method and system for predicting the harvest period of *Hymenochloa crus-galli* (a type of orchid) by integrating multi-source data. The method includes: aggregating meteorological elements, soil profile physicochemical indicators, and plant physiological signals of *Hymenochloa crus-galli* to obtain a spatiotemporally synchronized multidimensional information field; identifying the critical point where the growth pattern of *Hymenochloa crus-galli* undergoes a fundamental change, and dividing the complete growth cycle into multiple phenological stages; performing dimensionality reduction on the multidimensional data of the multiple phenological stages, and performing tensor synthesis on the trend characteristics of the dimensionality reduction results to obtain a growth trend vector; establishing a contribution quantification system and assigning contribution weights to the growth trend vector; correcting the growth trend vector, arranging the corrected results in chronological order, and considering them as the development trajectory of the entire life cycle to deduce the physiological maturity path; when the physiological maturity path reaches the historical optimal quality convergence interval of *Hymenochloa crus-galli*, it is determined as the optimal harvest window; this invention can improve the accuracy of *Hymenochloa crus-galli* harvest period prediction.

Description

Technical Field

[0001] This invention relates to the field of data reasoning technology, and in particular to a method and system for predicting the harvest period of *Hymenochloa crus-galli* by integrating multi-source data. Background Technology

[0002] Existing technologies for predicting the harvest period of *Illicium verum* (a type of orchid) have significant shortcomings in data collection and processing. They fail to simultaneously integrate meteorological elements, soil profile physicochemical indicators, and plant physiological signals to construct a spatiotemporally synchronized multidimensional information field. Analysis relies solely on single-type data or fragmented data that lacks spatiotemporal alignment, failing to comprehensively reflect the overall environment and physiological state of *Illicium verum* growth. Furthermore, they fail to identify critical points of growth pattern transitions through trend quantification and synergy strength assessment, dividing growth stages only according to fixed time points, ignoring the dynamic changes in *Illicium verum* growth. This results in a lack of scientific rigor and accuracy in phenological stage division, insufficient basic data support for subsequent growth status analysis, and directly impacts the reliability of prediction results.

[0003] Existing technologies have significant shortcomings in the analysis of growth status and harvest period determination for *Hymenochloa chinensis*. They fail to perform dimensionality reduction and tensor synthesis on multi-dimensional data from multiple phenological stages, simply integrating raw data or single-dimensional features. This fails to accurately extract core information about growth status, resulting in feature vectors with poor relevance and representativeness. Furthermore, they lack a quantitative system based on quality contribution, relying solely on fixed or empirical weights to assess growth status. This fails to reflect the differentiated impact of different phenological stages on final harvest quality, leading to significant discrepancies between the corrected growth trajectory and actual physiological maturity. Additionally, they do not construct convergence intervals using historical best quality data, relying solely on single physiological indicators or time experience to determine the harvest period. This makes it difficult to accurately capture the critical window for *Hymenochloa chinensis* to reach optimal quality, resulting in harvesting too early or too late, impacting the medicinal and economic value of the plant. Summary of the Invention

[0004] This invention provides a method and system for predicting the harvest period of *Hymenochloa crus-galli* by integrating multi-source data, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, this invention provides a method for predicting the harvest period of *Hymenochloa crus-galli* by fusing multi-source data, comprising:

[0006] S1. Synchronously sense and aggregate meteorological elements, soil profile physicochemical indicators and plant physiological signals of *Cymbidium goeringii* during its complete growth cycle to obtain a spatiotemporal synchronous multidimensional information field of *Cymbidium goeringii*.

[0007] S2. Based on the evolution law of the spatiotemporal synchronous multidimensional information field, identify the critical point at which the growth pattern of the Blood Leaf Orchid undergoes a fundamental change, and divide the complete growth cycle into multiple phenological stages of the Blood Leaf Orchid based on the critical point.

[0008] S3. Perform dimensionality reduction processing on the multi-dimensional data of the multi-phenological stages, and perform tensor synthesis on the situation characteristics of the dimensionality reduction results to obtain the growth situation vector of the multi-phenological stages.

[0009] S4. Based on the cumulative and specific contributions of the final harvest target in the *Hymenochloa chinensis*, establish a contribution quantification system for the *Hymenochloa chinensis* and assign contribution weights to the growth status vector.

[0010] S5. Apply the contribution weight to correct the growth trend vector, and arrange the corrected results in chronological order, which is regarded as the full life cycle development trajectory of the Blood Leaf Orchid, so as to deduce the physiological maturity path of the Blood Leaf Orchid.

[0011] S6. When the physiological maturity path reaches the historical best quality convergence interval of the Blood Leaf Orchid, it is determined as the best harvesting window for the Blood Leaf Orchid.

[0012] In a preferred embodiment, the synchronous sensing and aggregation of meteorological elements, soil profile physicochemical indicators, and plant physiological signals of *Hymenochloa chinensis* throughout its complete growth cycle to obtain a spatiotemporal synchronous multidimensional information field of *Hymenochloa chinensis* includes:

[0013] During the complete growth cycle of *Cymbidium goeringii*, meteorological elements, soil profile physicochemical indicators, and physiological signals of the plant were obtained.

[0014] Eliminate the temporal inconsistencies among the meteorological elements, the soil profile physicochemical indicators, and the plant's physiological signals to obtain the time-normalized data of the Blood Leaf Orchid.

[0015] In the spatial dimension, the geographical location information of the time-normalized data is registered to obtain the spatial alignment data of the Blood Leaf Orchid.

[0016] By integrating the attribute dimensions of different data sources in the spatial alignment data, a spatiotemporal synchronized multidimensional information field of the Bloodleaf Orchid is constructed.

[0017] In a preferred embodiment, the step of identifying the critical point at which the growth pattern of the *Cymbidium goeringii* undergoes a fundamental change based on the evolution law of the spatiotemporally synchronized multidimensional information field, and dividing the complete growth cycle into multiple phenological stages of the *Cymbidium goeringii* based on the critical point, includes:

[0018] The temporal changes of the meteorological elements, the soil profile physicochemical indicators, and the plant's physiological signals are quantified to obtain the trend quantification factor of the *Cymbidium goeringii*.

[0019] Based on the trend quantification factor, the coordination strength between different data sources in the spatiotemporal synchronized multidimensional information field is evaluated to obtain the trend coordination matrix of the Blood Leaf Orchid.

[0020] The dominant change pattern of the trend coordination matrix is ​​obtained by performing feature evolution analysis on the trend coordination matrix.

[0021] Based on the stability changes of the dominant change pattern, the critical point at which the growth pattern of the *Gynostemma pentaphyllum* undergoes a fundamental transformation is identified.

[0022] Based on the critical point, the complete growth cycle is divided into the initial phenological stage sequence of the *Hymenochloa chinensis*.

[0023] The phenological continuity of the initial phenological stage sequence is analyzed to adjust the boundaries of the initial phenological stage sequence, thereby obtaining the multiple phenological stages of the *Hymenochloa chinensis*.

[0024] In a preferred embodiment, the step of evaluating the coordination strength between different data sources in the spatiotemporally synchronized multidimensional information field based on the trend quantification factor to obtain the trend coordination matrix of *Hypericum erythrophyllum* includes:

[0025] The trend quantification factor is decomposed into a meteorological element trend layer, a soil profile trend layer, and a plant physiological trend layer.

[0026] The delayed cross-correlation of the first synergistic relationship strength between the meteorological element trend layer and the soil profile trend layer was analyzed to obtain the meteorological-soil synergistic relationship group of the *Hypericum erythrophyllum*.

[0027] The instantaneous coupling strength of the second synergistic relationship between the soil profile trend layer and the plant physiological trend layer was evaluated to obtain the soil-physiological synergistic relationship group of the *Hymenochloa chinensis*.

[0028] The strength of the third synergistic relationship between the plant's physiological trend layer and the meteorological element trend layer is transformed by nonlinear mutual information to obtain the physiological-meteorological synergistic relationship group of the *Hymenochloa chinensis*.

[0029] The meteorological-soil synergistic relationship group, the soil-physiological synergistic relationship group, and the physiological-meteorological synergistic relationship group are integrated in three dimensions to construct the trend synergistic matrix of *Hylocereus undatus*. The calculation formula of the trend synergistic matrix is ​​as follows:

[0030] ;

[0031] In the formula, The trend coordination matrix is... As the weighting factor for the delay effect, This refers to the meteorological-soil synergistic relationship group. For tensor product operations, This refers to the soil-physiological synergistic relationship group. For matrix direct sum operations, For immediate response weighting factors, This refers to the physiological-meteorological synergistic relationship group. This is a non-linear adaptive weighting factor.

[0032] In a preferred embodiment, the step of dimensionality reduction processing of the multi-dimensional data of the multi-phenological stages, and tensor synthesis of the trend features of the dimensionality reduction result to obtain the growth trend vector of the multi-phenological stages, includes:

[0033] The high-dimensional features of the multi-dimensional data in the multi-phenological stages are projected onto the low-dimensional manifold space to obtain the feature manifold representation of the multi-phenological stages.

[0034] Decouple the local features in the feature manifold representation to obtain the stable feature components of the feature manifold representation;

[0035] Reorganize the global features in the feature manifold representation to obtain the dynamic feature components of the feature manifold representation;

[0036] The stable feature components and the dynamic feature components are fused together at multiple scales to obtain the multi-level feature representation of the feature manifold representation.

[0037] Eliminate redundant information in the multi-level feature representation to determine the orthogonal feature base of the multi-level feature representation;

[0038] Based on the orthogonal characteristic basis, a characteristic tensor field for the multi-phenological stage is constructed, and tensor contraction is performed on the characteristic tensor field to determine the growth trend vector of the multi-phenological stage.

[0039] In a preferred embodiment, the step of establishing a contribution quantification system for the *Hymenochloa chinensis* based on the cumulative and specific contributions to the final harvest target, and assigning contribution weights to the growth status vector, includes:

[0040] The medicinal component content, biomass accumulation, and morphological development integrity of the *Hymenochloa chinensis* were extracted from historical harvesting data to form a quality characteristic set of the *Hymenochloa chinensis*.

[0041] The cumulative contribution of the multiple phenological stages to the set of quality characteristics is used as the first level of characterization, and the specific contribution of the multiple phenological stages to the key quality indicators in the set of quality characteristics is used as the second level of characterization to construct a multi-level contribution network for the Blood Leaf Orchid.

[0042] In the multi-level contribution network, the node centrality of the multi-phenological stage is evaluated, and the stage-based weights of the multi-phenological stage are determined.

[0043] Based on the similarity between the growth trend vector and the best harvested sample in the historical harvesting data, the basic weights of the stages are dynamically adjusted to obtain the contribution weights of the multi-phenological stages, and the contribution weights are assigned to the growth trend vector.

[0044] In a preferred embodiment, the application of the contribution weight to correct the growth status vector, and the arrangement of the corrected results in chronological order, are considered as the full life cycle development trajectory of the *Hymenochloa chinensis*, in order to deduce the physiological maturity path of the *Hymenochloa chinensis*, including:

[0045] The contribution weights are nonlinearly fused with the corresponding growth trend vectors to obtain the corrected growth trend vector of the *Hymenochloa chinensis*.

[0046] Arrange the corrected growth trend vectors in chronological order to obtain the growth trend sequence of the *Hypericum erythrophyllum*.

[0047] By eliminating the random fluctuation components of the growth status sequence, the full life cycle development trajectory of the *Hymenochloa chinensis* is obtained.

[0048] By performing trajectory curvature analysis on the entire life cycle development trajectory, the key turning points of trajectory changes in the entire life cycle development trajectory are obtained;

[0049] The key turning point region is mapped to the physiological maturity path of the Blood Leaf Orchid.

[0050] In a preferred embodiment, eliminating the random fluctuation component of the growth state sequence to obtain the full life cycle development trajectory of the *Hymenochloa chinensis* includes:

[0051] Multi-scale intrinsic mode decomposition is performed on the growth trend sequence to obtain the multi-level frequency domain features of the growth trend sequence.

[0052] Extract the significant modes related to growth rhythm and the interference modes related to environmental noise from the multi-level frequency domain features;

[0053] The saliency mode and the interference mode are phase synchronized to obtain the reconstructed reference mode of the growth state sequence;

[0054] The random fluctuation components of the reconstructed baseline mode are eliminated, and the processed results are subjected to trajectory evolution to obtain the full life cycle development trajectory of the Bloodleaf Orchid.

[0055] In a preferred embodiment, determining the optimal harvest window for the *Hymenochloa chinensis* when the physiological maturity path reaches the historical best quality convergence interval includes:

[0056] At the optimal harvest period in the historical harvest data, the physiological maturity characteristics of the *Hymenochloa chinensis* are used as the quality characteristic benchmark for the *Hymenochloa chinensis*.

[0057] The matching degree between the physiological maturity path and the quality characteristic benchmark is monitored in real time. When the physiological maturity path enters the quality convergence critical region of the Blood Leaf Orchid, the stable state of the physiological maturity path in the historical best quality convergence range is confirmed.

[0058] Based on the stable dwell state, the convergence strength and duration of the physiological maturity path are comprehensively judged to obtain the optimal harvesting window for the Blood Leaf Orchid.

[0059] To address the aforementioned problems, this invention also provides a system for predicting the harvest period of *Hymenochloa crus-galli* by integrating multi-source data. The system includes:

[0060] The multidimensional information field sensing module is used to synchronously sense and aggregate meteorological elements, soil profile physicochemical indicators and plant physiological signals of the Blood Leaf Orchid during its complete growth cycle, so as to obtain the spatiotemporal synchronous multidimensional information field of the Blood Leaf Orchid.

[0061] The phenological stage division module is used to identify the critical point at which the growth pattern of the *Cymbidium goeringii* undergoes a fundamental change based on the evolution law of the spatiotemporal synchronous multidimensional information field, and to divide the complete growth cycle into multiple phenological stages of the *Cymbidium goeringii* based on the critical point.

[0062] The growth trend vectorization module is used to perform dimensionality reduction processing on the multi-dimensional data of the multi-phenological stages, and to perform tensor synthesis on the trend features of the dimensionality reduction result to obtain the growth trend vector of the multi-phenological stages.

[0063] The contribution weight assignment module is used to establish a contribution quantification system for the *Hymenochloa chinensis* based on the cumulative and specific contributions formed by the final harvest target in the *Hymenochloa chinensis*, and to assign contribution weights to the growth status vector.

[0064] The life cycle trajectory extrapolation module is used to apply the contribution weight to correct the growth trend vector and arrange the corrected results in chronological order, which are regarded as the full life cycle development trajectory of the Blood Leaf Orchid, so as to extrapolate the physiological maturity path of the Blood Leaf Orchid.

[0065] The optimal harvesting decision module is used to determine the best harvesting window for the *Hymenochloa chinensis* when the physiological maturity path reaches the historical best quality convergence interval of the *Hymenochloa chinensis*.

[0066] Compared with the prior art, the present invention has the following beneficial effects:

[0067] 1. This invention lays a high-quality data foundation for predicting the harvest period of *Illicium lanceolatum* by integrating multi-source data and precise phenological classification. It simultaneously senses and aggregates meteorological elements, soil profile physicochemical indicators, and plant physiological signals throughout the complete growth cycle of *Illicium lanceolatum*. Through temporal consistency elimination, spatial coordinate registration, and attribute dimension integration, a spatiotemporally synchronized multidimensional information field is constructed to comprehensively capture the combined factors affecting the growth of *Illicium lanceolatum*. Through trend quantification, synergy intensity assessment, and feature evolution analysis, the critical points of growth pattern transformation are accurately identified. Combined with phenological continuity adjustment of stage boundaries, scientific multi-phenological stages are delineated, providing a precise stage division basis for subsequent growth status analysis.

[0068] 2. This invention significantly improves the accuracy and reliability of predicting the harvest period of *Hymenochloa chinensis* by utilizing refined growth status analysis and dynamic maturity projection. Tensor synthesis is performed on multi-phenological stage data after dimensionality reduction, fusing stable and dynamic features to construct a growth status vector. A contribution metric system is established based on historical quality data, dynamically assigning weights to the growth status vector. After correction, the vectors are arranged chronologically to form a full life-cycle development trajectory. The physiological maturity path is projected through trajectory curvature analysis. When the maturity path reaches the historical optimal quality convergence interval, the optimal harvest window is determined by combining convergence strength and persistence, ensuring a precise match between harvest timing and the optimal quality of *Hymenochloa chinensis*, effectively protecting its medicinal and economic value. Attached Figure Description

[0069] Figure 1 This is a flowchart illustrating a method for predicting the harvest period of *Hymenochloa crus-galli* by fusing multi-source data, provided in an embodiment of the present invention.

[0070] Figure 2 This is a functional module diagram of a multi-source data-integrated system for predicting the harvest period of *Hymenochloa crus-galli*, provided in an embodiment of the present invention.

[0071] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0072] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0073] This application provides a method for predicting the harvest period of *Gynostemma pentaphyllum* by integrating multi-source data. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for predicting the harvest period of *Gynostemma pentaphyllum* by integrating multi-source data can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0074] Reference Figure 1 The diagram shown is a flowchart illustrating a method for predicting the harvest period of *Hymenochloa crus-galli* (a type of orchid) by fusing multi-source data, according to an embodiment of the present invention. In this embodiment, the method for predicting the harvest period of *Hymenochloa crus-galli* by fusing multi-source data includes:

[0075] S1. Synchronously sense and aggregate meteorological elements, soil profile physicochemical indicators and plant physiological signals of *Cymbidium goeringii* during its complete growth cycle to obtain a spatiotemporal synchronous multidimensional information field of *Cymbidium goeringii*.

[0076] In this embodiment of the invention, the synchronous sensing and aggregation of meteorological elements, soil profile physicochemical indicators, and plant physiological signals of *Hylocereus undatus* during its complete growth cycle to obtain a spatiotemporal synchronous multidimensional information field of *Hylocereus undatus* includes:

[0077] During the complete growth cycle of *Cymbidium goeringii*, meteorological elements, soil profile physicochemical indicators, and physiological signals of the plant were obtained.

[0078] Eliminate the temporal inconsistencies among the meteorological elements, the soil profile physicochemical indicators, and the plant's physiological signals to obtain the time-normalized data of the Blood Leaf Orchid.

[0079] In the spatial dimension, the geographical location information of the time-normalized data is registered to obtain the spatial alignment data of the Blood Leaf Orchid.

[0080] By integrating the attribute dimensions of different data sources in the spatial alignment data, a spatiotemporal synchronized multidimensional information field of the Bloodleaf Orchid is constructed.

[0081] Throughout the complete growth cycle of *Cymbidium goeringii*, when acquiring meteorological elements, soil profile physicochemical indicators, and plant physiological signals, fixed meteorological monitoring equipment was deployed in the *Cymbidium goeringii* growing area to continuously collect information such as air temperature, air humidity, precipitation, and light intensity. The collection frequency was set according to the needs of different growth stages within the growth cycle to ensure coverage of all key periods, including budding, leaf expansion, flowering, and fruiting. Regarding soil profile physicochemical indicators, sampling points were set at different soil depths where the *Cymbidium goeringii* root system is distributed. Soil samples were collected periodically, and professional testing equipment was used to analyze indicators such as soil pH, organic matter content, nitrogen, phosphorus, and potassium nutrient concentrations, and soil moisture content. The time and corresponding soil depth of each sampling were recorded.

[0082] To target the plant's physiological signals, non-invasive sensors are attached to the leaves and stems of the blood-leaf orchid to monitor physiological parameters such as leaf chlorophyll content, transpiration rate, and stem diameter changes in real time. This ensures that the collected signals reflect the plant's real-time growth status and ultimately obtains three types of raw data covering the entire growth cycle.

[0083] To eliminate temporal inconsistencies among meteorological elements, soil profile physicochemical indicators, and plant physiological signals, and to obtain time-normalized data for *Gynostemma pentaphyllum*, a unified time reference is first established, using the collection interval of plant physiological signals as the standard, as this interval typically has the highest collection frequency and more accurately reflects the plant's real-time state. For meteorological element data, if the collection interval is greater than the unified time reference, linear interpolation is used to calculate the estimated meteorological element value corresponding to the reference time point based on the meteorological element values ​​of two adjacent collection time points; if the collection interval is less than the unified time reference, the meteorological element data closest to the reference time point is selected as the value for that reference time point.

[0084] For soil profile physicochemical index data, since the collection frequency is usually low, the most recently collected physicochemical index data is directly used if the time difference between the collection time and the baseline time point is small; if the time difference is large, the soil profile physicochemical index data is reasonably extrapolated by considering the changing trends of meteorological elements during that period, such as the impact of precipitation on soil moisture content and the impact of temperature on soil nutrient transformation, to supplement the values ​​at the baseline time point. After the above processing, all three types of data have corresponding values ​​at each baseline time point, resulting in time-normalized data.

[0085] In terms of spatial dimension, when registering the geographic location information of time-normalized data to obtain the spatial alignment data of *Hymenochloa chinensis*, a unified spatial coordinate system is first established for the *Hymenochloa chinensis* growing area to determine the precise geographic location coordinates of each *Hymenochloa chinensis* plant, including longitude, latitude, and altitude. For meteorological element data, the specific location coordinates of meteorological monitoring equipment in this spatial coordinate system are recorded, and the meteorological element data of each time reference point is associated with the corresponding equipment location coordinates to ensure that the meteorological data can accurately correspond to the specific spatial location of the *Hymenochloa chinensis* growing area.

[0086] For soil profile physicochemical index data, the location coordinates of each soil sampling point in the spatial coordinate system and the sampling depth are recorded. The processed soil physicochemical index data are then bound to the sampling point location and soil depth information, enabling the soil data to correspond to specific spatial locations and soil layers. For plant physiological signal data, the geographical coordinates of the plants to which the sensors are attached are directly associated with the physiological signal data. Subsequently, using the plant geographical coordinates as the core, meteorological element data, soil profile physicochemical index data, and plant physiological signal data within the same spatial range are matched to ensure that different types of data correspond one-to-one in spatial location, resulting in spatially aligned data.

[0087] When constructing a spatiotemporally synchronized multidimensional information field for *Illicium lanceolatum* by integrating the attribute dimensions of different data sources in spatially aligned data, the attribute dimensions of each data source in the spatially aligned data are first sorted out. The attribute dimensions of meteorological element data include air temperature and air humidity, the attribute dimensions of soil profile physicochemical index data include soil pH and organic matter content, and the attribute dimensions of plant physiological signal data include leaf chlorophyll content and transpiration rate. Based on the time reference point and spatial location coordinates, data with different attribute dimensions corresponding to the same time and spatial location are integrated to form a dataset containing time, space, and multiple attribute dimensions.

[0088] During the integration process, a unified data format was used to record each data entry, ensuring that each entry included a timestamp, spatial coordinates, and specific values ​​for each attribute dimension, without any data format conflicts or missing attributes. Through this integration method, scattered single-dimensional data were merged into a whole that simultaneously reflects the temporal changes, spatial distribution, and multiple environmental and physiological attributes of *Illicium verum* during its growth, ultimately constructing a spatiotemporally synchronized multidimensional information field for *Illicium verum*.

[0089] The beneficial effects include the simultaneous acquisition of meteorological elements, soil profile physicochemical indicators, and plant physiological signals. These three types of data correspond to the external environment, soil conditions, and the plant's own physiological state, respectively, comprehensively covering the key dimensions affecting growth. Compared to a single data source, multi-source data acquisition avoids information bias and ensures that subsequent analysis comprehensively reflects the overall conditions for the growth of *Hymenochloa chinensis*, providing rich and comprehensive raw data support for the construction of a spatiotemporally synchronized multidimensional information field.

[0090] By unifying the time reference, adjustments were made to three types of data collected at different frequencies, ensuring that meteorological elements, soil profile physicochemical indicators, and plant physiological signals corresponded at the same time points. This processing resolved the temporal misalignment problem caused by differences in collection cycles among multi-source data, ensuring that time-standardized data accurately reflects the state of various influencing factors at the same moment, laying a precise temporal foundation for subsequent spatial registration and attribute integration.

[0091] A unified spatial coordinate system was established, binding time-normalized data with spatial information such as specific geographical locations and soil sampling depths. This enabled meteorological, soil, and plant physiological data to be accurately mapped to the specific spatial location of *Hylocereus undatus* growth. Spatially aligned data eliminated spatial misalignment interference from different data sources, ensuring the correlation and matching degree of various data in the spatial dimension. This allowed subsequent integration to accurately reflect the growth correlation information of *Hylocereus undatus* under specific spatial conditions.

[0092] By integrating attribute dimensions from different data sources in spatially aligned data according to time and spatial coordinates, a unified dataset encompassing multiple dimensions such as time, space, environment, and physiology is formed. This spatiotemporally synchronized multidimensional information field breaks down the isolation of single data sources, achieving deep correlation among multiple data sources. It comprehensively and coherently presents the dynamic changes and interactions of various factors during the growth of *Illicium verum*, providing high-quality comprehensive data support for subsequent critical point identification and phenological stage division, significantly improving the scientific rigor and accuracy of harvest period prediction.

[0093] S2. Based on the evolution law of the spatiotemporal synchronous multidimensional information field, identify the critical point at which the growth pattern of the Blood Leaf Orchid undergoes a fundamental change, and divide the complete growth cycle into multiple phenological stages of the Blood Leaf Orchid based on the critical point.

[0094] In this embodiment of the invention, the step of identifying the critical point at which the growth pattern of the *Cymbidium goeringii* undergoes a fundamental change based on the evolution law of the spatiotemporally synchronized multidimensional information field, and dividing the complete growth cycle into multiple phenological stages of the *Cymbidium goeringii* based on the critical point, includes:

[0095] The temporal changes of the meteorological elements, the soil profile physicochemical indicators, and the plant's physiological signals are quantified to obtain the trend quantification factor of the *Cymbidium goeringii*.

[0096] Based on the trend quantification factor, the coordination strength between different data sources in the spatiotemporal synchronized multidimensional information field is evaluated to obtain the trend coordination matrix of the Blood Leaf Orchid.

[0097] The dominant change pattern of the trend coordination matrix is ​​obtained by performing feature evolution analysis on the trend coordination matrix.

[0098] Based on the stability changes of the dominant change pattern, the critical point at which the growth pattern of the *Gynostemma pentaphyllum* undergoes a fundamental transformation is identified.

[0099] Based on the critical point, the complete growth cycle is divided into the initial phenological stage sequence of the *Hymenochloa chinensis*.

[0100] The phenological continuity of the initial phenological stage sequence is analyzed to adjust the boundaries of the initial phenological stage sequence, thereby obtaining the multiple phenological stages of the *Hymenochloa chinensis*.

[0101] The step of evaluating the coordination strength between different data sources in the spatiotemporally synchronized multidimensional information field based on the trend quantification factor to obtain the trend coordination matrix of the *Hymenochloa chinensis* includes:

[0102] The trend quantification factor is decomposed into a meteorological element trend layer, a soil profile trend layer, and a plant physiological trend layer.

[0103] The delayed cross-correlation of the first synergistic relationship strength between the meteorological element trend layer and the soil profile trend layer was analyzed to obtain the meteorological-soil synergistic relationship group of the *Hypericum erythrophyllum*.

[0104] The instantaneous coupling strength of the second synergistic relationship between the soil profile trend layer and the plant physiological trend layer was evaluated to obtain the soil-physiological synergistic relationship group of the *Hymenochloa chinensis*.

[0105] The strength of the third synergistic relationship between the plant's physiological trend layer and the meteorological element trend layer is transformed by nonlinear mutual information to obtain the physiological-meteorological synergistic relationship group of the *Hymenochloa chinensis*.

[0106] The meteorological-soil synergistic relationship group, the soil-physiological synergistic relationship group, and the physiological-meteorological synergistic relationship group are integrated in three dimensions to construct the trend synergistic matrix of *Hylocereus undatus*. The calculation formula of the trend synergistic matrix is ​​as follows:

[0107] ;

[0108] In the formula, The trend coordination matrix is... As the weighting factor for the delay effect, This refers to the meteorological-soil synergistic relationship group. For tensor product operations, This refers to the soil-physiological synergistic relationship group. For matrix direct sum operations, For immediate response weighting factors, This refers to the physiological-meteorological synergistic relationship group. This is a non-linear adaptive weighting factor.

[0109] The temporal changes of meteorological elements, soil profile physicochemical indicators, and plant physiological signals were quantified to obtain trend quantification factors for *Gynostemma pentaphyllum*. Time series data of each of the three types of signals throughout the complete growth cycle were analyzed one by one to observe the trend of each signal over time, including increases, decreases, stabilization, or fluctuations. For each signal's time series, the rate of change of the signal in different time periods was determined by calculating the amplitude and direction of change of data at adjacent time points. Then, considering the importance of the signal in the growth cycle, the rate of change was weighted, and the processed results were integrated into quantitative indicators that intuitively reflect the temporal trend of various signals. These quantitative indicators are the trend quantification factors, ensuring that each factor accurately corresponds to the temporal change characteristics of a specific type of signal.

[0110] Based on trend quantification factors, the synergy strength between different data sources in a spatiotemporally synchronized multidimensional information field is evaluated to obtain the trend synergy matrix for *Illicium lanceolatum*. It is clearly defined that different data sources correspond to meteorological elements, soil profile physicochemical indicators, and plant physiological signals, respectively, and each type of data source has a corresponding trend quantification factor. By comparing the changing patterns of trend quantification factors from different data sources, their synchronicity and correlation in the time dimension are analyzed. For example, if the trend quantification factor of meteorological elements increases within a certain time period, and the trend quantification factor of soil profile physicochemical indicators also shows a change in the same direction, it indicates strong synergy between the two. Using fixed evaluation criteria, the degree of synergy between any two types of data sources is scored. All scoring results are arranged in a preset row and column order to form a matrix. Each element in the matrix represents the synergy strength between the corresponding two types of data sources; this matrix is ​​the trend synergy matrix.

[0111] When performing feature evolution analysis on the trend coordination matrix to obtain its dominant change pattern, the changes in the trend coordination matrix are continuously tracked throughout its complete growth cycle, recording the numerical fluctuations of each element in the matrix over time and the overall structural adjustments of the matrix. By observing the elements in the matrix with the most significant numerical changes and the data source combinations they represent, the core data source relationships that play a decisive role in the overall changes of the matrix are identified. Combining the changing patterns of these core relationships, the main change characteristics of the trend coordination matrix in different time periods are summarized. These characteristics together constitute the dominant change pattern of the trend coordination matrix, which reflects the core evolutionary laws of the coordination relationships between different data sources.

[0112] Based on the stability changes of the dominant change pattern, when identifying the critical point of a fundamental shift in the growth pattern of *Illicium lanceolatum*, continuously monitor the stable state of the dominant change pattern within the growth cycle to determine whether it maintains a consistent pattern and characteristics. When the dominant change pattern undergoes a sudden and irreversible change, such as the previously stable core data source relationship being replaced by a new relationship, or the change pattern shifting from an upward trend to a downward trend that cannot be reversed, it indicates a fundamental adjustment in the synergistic relationship between different data sources. The time point corresponding to this adjustment is precisely the key point of a fundamental shift in the growth pattern of *Illicium lanceolatum*. This time point is determined as the critical point to ensure that the critical point accurately corresponds to the essential change in the growth pattern.

[0113] When dividing the complete growth cycle into the initial phenological stage sequence of *Cymbidium goeringii* based on critical points, all identified critical points serve as boundaries, and the complete growth cycle is divided into several consecutive time periods in chronological order. Each time period begins with one critical point and ends with the next critical point, ensuring that the dominant change pattern remains stable within each time period and that the growth pattern does not undergo fundamental changes. Each time period is assigned a corresponding time range identifier, and these time periods are arranged sequentially to form the initial phenological stage sequence. Each stage in the sequence corresponds to a period in the growth cycle with a stable growth pattern.

[0114] To analyze the phenological continuity of the initial phenological stage sequence and adjust its boundaries to obtain the multi-phenological stages of *Cymbidium goeringii*, the connection between adjacent stages in the initial phenological stage sequence is examined one by one. This involves observing whether the growth characteristics of the previous stage can naturally transition to the next stage, and whether the growth characteristics of the next stage are based on the evolution of the previous stage, ensuring that there are no logical breaks or conflicting characteristics between stages. If the continuity of phenological characteristics between two adjacent stages is found to be poor—for example, if the core growth indicators of the previous stage are not reasonably continued in the subsequent stage, or if unrelated new characteristics appear—the boundary critical points of the two stages are adjusted based on the natural evolutionary logic of phenological characteristics. This ensures that the adjusted stage boundaries better reflect the continuous change pattern of phenological characteristics. After analyzing the continuity and adjusting the boundaries of all adjacent stages, a multi-phenological stage structure that is complete, logically coherent, and accurately reflects the growth and evolution process of *Cymbidium goeringii* is finally formed.

[0115] When decomposing the trend quantification factor into meteorological element trend layer, soil profile trend layer, and plant physiological trend layer, the original data source attributes corresponding to each type of data in the trend quantification factor are first clarified. Quantitative indicators related to meteorological elements are then selected and integrated according to their temporal change characteristics to form the meteorological element trend layer, which fully preserves the temporal change quantitative information of meteorological elements. Quantitative indicators related to soil profile physicochemical indicators are then selected and categorized and integrated according to the attribute associations and temporal change patterns of soil data to form the soil profile trend layer, ensuring the completeness and relevance of soil-related quantitative information. Quantitative indicators related to plant physiological signals are then selected and organized according to the logical associations and temporal evolution order of plant physiological parameters to form the plant physiological trend layer, enabling the orderly presentation of plant physiological quantitative information.

[0116] This study analyzes the delayed cross-correlation of the first synergistic relationship strength between the meteorological element trend layer and the soil profile trend layer, obtaining a meteorological-soil synergistic relationship group for *Illicium verum*. Using a time axis as a baseline, the changes in quantitative indicators of the meteorological element trend layer and the soil profile trend layer are compared segment by segment. For the changes in the quantitative indicators of meteorological elements in each time period, the response changes of the quantitative indicators of the soil profile in subsequent time intervals are tracked to determine the degree of correlation and time delay characteristics of the two changes. By analyzing the consistency and strength of this delayed response, the synergistic relationship between the two is quantified. The synergistic strength data of different time periods are organized in chronological order to form a meteorological-soil synergistic relationship group containing delayed correlation information and synergistic strength quantitative values.

[0117] To assess the instantaneous coupling strength of the second synergistic relationship between the soil profile trend layer and the plant physiological trend layer, and to obtain the soil-physiological synergistic relationship group for *Spatholobus suberectus*, the focus is on the quantitative index values ​​of the soil profile trend layer and the plant physiological trend layer at the same time point. The analysis examines whether the changing directions of the two types of quantitative indicators are consistent and whether the magnitudes of change show correlation at the same time point. By judging the degree of synchronous change of the two at the instantaneous time point, the tightness of instantaneous coupling is quantified. The quantitative values ​​of the instantaneous coupling strength at each time point are arranged in chronological order to form a soil-physiological synergistic relationship group that reflects the tightness of soil-plant physiological synergy at different times.

[0118] Nonlinear mutual information transformation was performed on the strength of the third synergistic relationship between the plant's physiological trend layer and the meteorological element trend layer to obtain the physiological-meteorological synergistic relationship group of *Cymbidium goeringii*. The time series of quantitative indicators of the plant's physiological trend layer and the meteorological element trend layer throughout the complete growth cycle were then analyzed. The nonlinear correlation patterns between the two types of sequences were analyzed to uncover hidden correlation information independent of linear relationships. By quantifying the information transmission volume and correlation tightness of this nonlinear correlation, it was transformed into a synergistic strength quantification value on a unified scale. These quantification values ​​were integrated chronologically to form a physiological-meteorological synergistic relationship group that can characterize the nonlinear synergistic relationship between plant physiology and meteorological elements.

[0119] To construct a trend synergy matrix for *Illicium lanceolatum* by integrating the meteorological-soil synergy group, the soil-physiological synergy group, and the physiological-meteorological synergy group in three dimensions, the three dimensions of the matrix were first determined to correspond to the three types of synergy groups, with each dimension's coordinate node corresponding to a reference time point on the time axis. The quantified synergy intensity values ​​for each time point in the meteorological-soil synergy group were filled into the meteorological-soil dimension position of the corresponding time node; the quantified values ​​for the soil-physiological synergy group were filled into the soil-physiological dimension position of the corresponding time node; and the quantified values ​​for the physiological-meteorological synergy group were filled into the physiological-meteorological dimension position of the corresponding time node. This ensures that the quantified values ​​of the three types of synergy at each time node can be accurately located in the matrix, forming a three-dimensional matrix that comprehensively reflects the synergy intensity between different time periods and different data sources. This matrix is ​​the trend synergy matrix.

[0120] The meteorological-soil synergistic relationship group originates from the delayed cross-correlation analysis of the meteorological element trend layer and the soil profile trend layer. Based on the time axis, the changes in the quantitative indicators of the two types of trend layers are compared segment by segment. The delayed response of soil profile indicators after changes in meteorological elements is tracked, the synergistic relationship between the two is quantified and organized in chronological order, forming a relationship group containing delayed correlation information and quantitative values ​​of synergistic strength.

[0121] The soil-physiological synergy group is derived from the instantaneous coupling strength assessment of the soil profile trend layer and the plant physiological trend layer. It focuses on the quantitative indicators of the two types of trend layers at the same time point, analyzes the degree of instantaneous synchronous change and quantifies the tightness of coupling, and arranges the quantitative values ​​of each time point in order to form this relationship group.

[0122] The physiological-meteorological synergistic relationship group originates from the nonlinear mutual information transformation between the plant physiological trend layer and the meteorological element trend layer. By sorting out the time series of the two types of trend layers, the nonlinear hidden correlations are explored and the information transmission volume and correlation tightness are quantified. After being transformed into a synergistic strength quantification value on a unified scale, the relationship group is formed by integrating them over time.

[0123] The delayed effect weighting factor is pre-set based on the delayed characteristics of the impact of meteorological elements on the soil environment. By analyzing the delayed duration and influence intensity of the meteorological-soil synergy in a large number of *Hydrocotyle sibthorpioides* growth samples, weights are assigned to the tensor product calculation results of the meteorological-soil synergy group and the soil-physiological synergy group to ensure that the delayed effect is reasonably reflected in the matrix.

[0124] The immediate response weighting factor is pre-set based on the immediate characteristics of the impact of soil environmental changes on plant physiological state. According to the degree of influence of the instantaneous coupling strength of soil-physiological synergy on plant growth, the tensor product calculation results of soil-physiological synergy group and physiological-meteorological synergy group are assigned weights to highlight the role of immediate response.

[0125] The nonlinear adaptation weighting factor is pre-set based on the nonlinear adaptation law of plant physiological state to changes in meteorological elements. It combines the nonlinear correlation strength of physiological-meteorological synergy and its impact on overall growth to assign weights to the tensor product calculation results of physiological-meteorological synergy group and meteorological-soil synergy group, thus adapting to the nonlinear synergy characteristics.

[0126] Tensor product operation is used to expand the dimension and fuse information of two types of cooperative relationship groups. It correlates the quantized value of each cooperative strength of the former relationship group with the corresponding quantized value of the latter relationship group to generate a higher-dimensional joint feature matrix, which fully preserves the cooperative information and correlation characteristics of the two relationship groups.

[0127] The matrix direct sum operation is used to integrate three feature matrices after tensor product operation and weight allocation. According to the cooperative relationship type corresponding to the matrix dimension, the elements of each matrix are filled into the corresponding positions of the overall matrix to ensure that different types of cooperative information do not conflict with each other and are fully presented in the matrix, ultimately forming a unified comprehensive matrix.

[0128] The formula signifies that, through the synergistic effects of weight allocation, tensor product fusion, and direct sum integration, it comprehensively integrates information from three types of synergistic relationship groups, generating a trend synergy matrix that accurately characterizes the multi-dimensional synergistic strength among meteorological elements, soil profile physicochemical indicators, and plant physiological signals. In the calculation, the meteorological-soil synergistic relationship group and the soil-physiological synergistic relationship group are first subjected to tensor product operation to expand the dimensions and fuse the synergistic information of both. Then, the calculation results are weighted and adjusted using a delay effect weighting factor to highlight the influence of delayed synergy, resulting in the first part of the feature matrix.

[0129] Simultaneously, tensor product operations are performed on the soil-physiological synergy group and the physiological-meteorological synergy group to integrate the correlation information of instantaneous synergy and nonlinear synergy. After weighting correction by the immediate response weight factor, the second part of the feature matrix is ​​obtained. Then, tensor product operations are performed on the physiological-meteorological synergy group and the meteorological-soil synergy group to integrate the synergistic information of nonlinear adaptation and delayed effects. After adjustment by the nonlinear adaptation weight factor, the third part of the feature matrix is ​​obtained.

[0130] Finally, the three weighted feature matrices are integrated through matrix summation operations, and matrix elements are filled in according to the type of synergy relationship to form a trend synergy matrix that can comprehensively reflect the strength of different synergy relationships and cover multi-dimensional correlation information, providing accurate matrix data support for subsequent feature evolution analysis and critical point identification.

[0131] The beneficial effect is that by quantifying the temporal changes of meteorological elements, soil profile physicochemical indicators, and plant physiological signals, the dynamic changes of these three types of data are transformed into intuitive trend quantification factors, accurately capturing the core patterns of increase, decrease, and stabilization of each data point over time. This process avoids the problem of ambiguous change characteristics caused by fluctuations in the original data, providing a clear and quantifiable analytical basis for subsequent assessment of the synergistic strength of different data sources, ensuring that subsequent operations can focus on the essential changes in the data.

[0132] Based on trend quantification factors, the synergy strength between different data sources is assessed, forming a trend synergy matrix. Each element in the matrix precisely corresponds to the degree of synergy between two types of data sources, fully presenting the dynamic relationship between meteorological, soil, and plant data. Compared to analyzing a single data source, this matrix can intuitively reflect the interactive effects of multi-source data, such as the driving effect of meteorological element changes on soil indicators and the relationship between soil conditions and plant physiology, providing key support for exploring the intrinsic driving factors of growth pattern transformation.

[0133] By performing feature evolution analysis on the trend coordination matrix, the dominant change pattern that plays a decisive role in the overall change of the matrix can be identified. This pattern can extract the core evolutionary law of the synergistic relationship of multi-source data and eliminate the interference of secondary fluctuations on the analysis. By focusing on the dominant change pattern, the key logic of the interaction of various factors in the growth process of *Illicium lanceolatum* can be clearly grasped, avoiding deviation in the analysis direction due to the complexity of data, and providing a clear target for accurately identifying the nodes of growth pattern transformation.

[0134] By identifying critical points based on the stability changes of the dominant growth pattern, a sudden and irreversible change in the dominant pattern is determined as the time node for a fundamental shift in the growth pattern. This method can accurately capture the key moment when the growth state transitions from one stable pattern to another. Compared with the traditional method of dividing stages according to fixed time, the division based on critical points is more in line with the dynamic laws of *Illicium erinaceus* growth, providing a scientific and objective boundary basis for subsequent phenological stage division.

[0135] First, the initial phenological stage sequence is divided based on the critical point. Then, the stage boundaries are adjusted by analyzing the continuity of phenology to ensure that the growth characteristics of adjacent stages can transition naturally without logical breaks. This process not only preserves the essential change in growth pattern reflected by the critical point, but also makes up for the stage connection problems that may exist in the initial division by adjusting the boundaries. The resulting multi-phenological stages can completely and coherently reflect the entire process of *Illicium verum* from the early stage of growth to maturity, providing an accurate stage division basis for subsequent growth status analysis and harvest period prediction.

[0136] The trend quantification factor is decomposed into three layers based on data source attributes: meteorological element trend layer, soil profile trend layer, and plant physiological trend layer. Each trend layer retains the temporal variation characteristics of its corresponding data source separately, avoiding confusion between features of different data types. This hierarchical processing clearly defines the trend attributes of each data source, providing a clear analytical object for subsequent targeted analysis of the synergistic relationships between different data sources, and ensuring that the assessment of synergistic strength does not deviate from the essential characteristics of each data source.

[0137] Differential analysis methods are employed to analyze the correlation characteristics of different data source combinations. For the meteorological-soil combination, delayed cross-correlation is analyzed to accurately capture the lag response patterns of soil indicators after changes in meteorological elements. For the soil-physiological combination, instantaneous coupling strength is assessed to reflect the immediate impact of soil environmental changes on plant physiological states. For the physiological-meteorological combination, nonlinear mutual information transformation is performed to adapt to the complex nonlinear adaptation relationship of plant physiology to meteorological elements. These three differential analysis methods are matched to the inherent correlation characteristics of the three types of data source combinations, ensuring that the resulting meteorological-soil synergistic relationship groups, soil-physiological synergistic relationship groups, and physiological-meteorological synergistic relationship groups accurately reflect the actual synergistic state, avoiding the bias in synergistic relationship representation caused by a single analysis method.

[0138] The tensor product results of meteorological-soil, soil-physiological, and physiological-meteorological relationships are weighted using delayed effect weighting, immediate response weighting, and nonlinear adaptation weighting, respectively. These weighting factors are set based on the degree of influence of each synergistic relationship on the growth of *Illicium lanceolatum*. For example, the delayed effect weighting emphasizes the long-term, lagged impact of meteorology on soil, the immediate response weighting emphasizes the rapid effect of soil on the plant, and the nonlinear adaptation weighting adapts to the complex relationship between physiological and meteorological factors. This weighting process strengthens the information of key synergistic relationships in the trend synergy matrix while reasonably weakening the influence of secondary synergistic relationships, ensuring that the matrix prioritizes reflecting the synergistic information that plays a dominant role in the growth pattern.

[0139] Tensor product operations expand the dimensions and deepen the associations of two types of synergistic relationship groups, fully preserving the detailed information and correlation characteristics of both types of synergistic relationships. Matrix direct sum operations integrate the three weighted tensor product results into a unified matrix, ensuring that different types of synergistic information are presented without conflict and in a complete manner within the matrix. By combining these two operations, three scattered synergistic relationship groups are merged into a trend synergistic matrix that comprehensively reflects the multi-dimensional synergistic strength among meteorology, soil, and plants, providing structurally complete and informationally accurate matrix data support for subsequent feature evolution analysis and critical point identification.

[0140] S3. Perform dimensionality reduction processing on the multi-dimensional data of the multi-phenological stages, and perform tensor synthesis on the situation characteristics of the dimensionality reduction results to obtain the growth situation vector of the multi-phenological stages.

[0141] In this embodiment of the invention, the step of performing dimensionality reduction processing on the multi-dimensional data of the multi-phenological stages and tensor synthesis on the trend features of the dimensionality reduction results to obtain the growth trend vector of the multi-phenological stages includes:

[0142] The high-dimensional features of the multi-dimensional data in the multi-phenological stages are projected onto the low-dimensional manifold space to obtain the feature manifold representation of the multi-phenological stages.

[0143] Decouple the local features in the feature manifold representation to obtain the stable feature components of the feature manifold representation;

[0144] Reorganize the global features in the feature manifold representation to obtain the dynamic feature components of the feature manifold representation;

[0145] The stable feature components and the dynamic feature components are fused together at multiple scales to obtain the multi-level feature representation of the feature manifold representation.

[0146] Eliminate redundant information in the multi-level feature representation to determine the orthogonal feature base of the multi-level feature representation;

[0147] Based on the orthogonal characteristic basis, a characteristic tensor field for the multi-phenological stage is constructed, and tensor contraction is performed on the characteristic tensor field to determine the growth trend vector of the multi-phenological stage.

[0148] Projecting high-dimensional features of multi-dimensional data from multiple phenological stages onto a low-dimensional manifold space yields a characteristic manifold representation of these stages. This process clarifies that the multi-dimensional data includes various attributes such as meteorological elements, soil profile physicochemical indicators, and plant physiological signals, constituting a high-dimensional feature set. Using a fixed projection method, the core relationships and key change patterns of various data types within the high-dimensional features are preserved, while irrelevant secondary information is removed, mapping the high-dimensional features to the low-dimensional manifold space. During this mapping process, it is ensured that the feature distribution in the low-dimensional space accurately reproduces the essential structure of the high-dimensional features, with each low-dimensional feature point corresponding to a set of associated data within the high-dimensional features. Ultimately, this results in a characteristic manifold representation that characterizes the core features of the high-dimensional data from multiple phenological stages.

[0149] When decoupling local features from the characteristic manifold representation to obtain stable feature components, each local region in the representation is analyzed individually to identify stable feature information that remains unchanged over time. This information is unaffected by short-term fluctuations within phenological stages and represents the inherent characteristics of that local region. By separating these stable local features from other dynamically changing features, all stable local features are integrated according to their correlations to form stable feature components of the characteristic manifold representation. These components reflect the core characteristic attributes that remain unchanged across multiple phenological stages.

[0150] When recombining global features in the characteristic manifold representation to obtain the dynamic feature components, global features that span different local regions and reflect the changing patterns of phenological stages are extracted from the overall perspective of the characteristic manifold representation. The changing trends and evolution patterns of these global features over time are analyzed, and they are recombined according to chronological order and change logic, so that the recombined features can clearly present the dynamic changes of multiple phenological stages. The integrated features focus on the dynamic evolution information within phenological stages, forming the dynamic feature components of the characteristic manifold representation, which can accurately reflect the changing characteristics of multiple phenological stages.

[0151] By performing multi-scale feature cross-fusion of stable and dynamic feature components to obtain a feature manifold representation, multiple different scale levels are set, each corresponding to a different feature analysis granularity. At each scale level, the core features in the stable feature components are cross-compared and information is fused with the changing features in the dynamic feature components. This allows stable features to provide foundational support for dynamic features, while dynamic features supplement the stable features with information about changes. By repeating this fusion process at different scale levels, the fusion results of each level are integrated to form a multi-level feature representation that covers different granularities and possesses both stable attributes and dynamic changes, comprehensively preserving the core information of multi-dimensional data.

[0152] To eliminate redundant information in multi-level feature representations and determine the orthogonal feature basis, each feature term in the multi-level feature representation is compared one by one to identify redundant features that are repetitive and do not contribute additionally to the feature representation. These redundant features are then removed using a fixed filtering method, retaining key feature terms that are independent and can represent data features from different perspectives. These key feature terms are then orthogonalized so that the processed feature terms are perpendicular to each other and have no information overlap. Each feature term can independently represent the core information of a certain dimension, ultimately forming the orthogonal feature basis of the multi-level feature representation, providing a concise and effective foundation for the subsequent construction of the feature tensor field.

[0153] Based on orthogonal characteristic bases, a feature tensor field for multiple phenological stages is constructed. This feature tensor field is then subjected to tensor contraction to determine the growth trend vector for each phenological stage. Using the orthogonal characteristic bases as a foundation, and according to the temporal order and feature dimensions of the multiple phenological stages, the feature data of each phenological stage are associated with the orthogonal characteristic bases, constructing a feature tensor field containing multi-dimensional information including time, space, and features. Each tensor element in this tensor field corresponds to core data of a specific phenological stage and a specific feature dimension. Through tensor contraction, redundant dimensions in the feature tensor field are compressed, retaining key dimensional information that reflects the growth trend, transforming the high-dimensional feature tensor field into a low-dimensional vector form. During the contraction process, it is ensured that the vector accurately summarizes the core growth characteristics and overall change trends of the multiple phenological stages, ultimately yielding the growth trend vector for each stage.

[0154] The beneficial effects are that when projecting high-dimensional features of multi-phenological, multi-dimensional data onto a low-dimensional manifold space, redundant information with weak correlation to the growth status of *Illicium lanceolatum* is eliminated from the high-dimensional data, such as invalid data fluctuations caused by short-term environmental noise. At the same time, the core correlations and key change patterns among meteorological, soil, and plant physiological data are fully preserved. The generated feature manifold representation accurately restores the essential features of the high-dimensional data in a low-dimensional form, avoiding the computational inefficiency caused by the curse of dimensionality and ensuring that subsequent feature processing always revolves around the core growth information, laying a concise and high-quality foundation for the construction of the growth status vector.

[0155] When decoupling local features from the feature manifold representation, stable features that are unaffected by short-term environmental disturbances and do not fluctuate over time are accurately separated. These include basic leaf morphology at specific phenological stages and the inherent rhythm of root development. These features are inherent attributes of *Cymbidium goeringii* during its growth process and constitute the core benchmark of its growth status. Obtaining stable feature components avoids interference from dynamic changes in the judgment of core growth attributes, providing a reliable reference for subsequent fusion of dynamic features and ensuring that the growth status vector truly reflects the plant's essential growth state.

[0156] When recombining global features in the manifold representation, a holistic perspective across multiple phenological stages is adopted to analyze feature information spanning different local regions. This includes changes in the accumulation rate of medicinal components throughout the entire cycle and the phased trends in biomass growth. These features are then integrated chronologically and logically to form a dynamic feature component. This component accurately captures the growth characteristics of *Illicium verum* at different phenological stages, supplementing the dynamic evolutionary information that stable features cannot cover. This allows the representation of growth status to include both inherent attributes and phased change patterns, achieving a comprehensive characterization of the growth state.

[0157] When performing multi-scale feature cross-fusion of stable and dynamic feature components, different granularity analysis scales are set, such as microscopic diurnal variations of physiological indicators and macroscopic phenological stage transition characteristics. At each scale level, a deep integration of stable attributes and dynamic changes is achieved. Stable features provide the foundational support for dynamic features, while dynamic features supplement the stable features with details of change. This multi-level feature representation encompasses multi-dimensional information from local to global, and from static to dynamic, avoiding the feature bias caused by single-scale analysis and providing rich and comprehensive material for subsequent optimization of the feature base.

[0158] When eliminating redundant information in multi-level feature representations, features that are repetitive or do not contribute additionally to the characterization of growth status are removed, such as highly similar meteorological and soil indirect influence features, while retaining independent key features. Then, orthogonalization is applied to ensure that the remaining features are perpendicular to each other and have no information overlap. Each feature can independently represent core growth information in a specific dimension, such as reflecting the accumulation of medicinal components, biomass growth, or morphological development. The final orthogonal feature basis carries multi-dimensional key information with minimal redundancy, providing accurate and efficient feature support for the subsequent construction of the feature tensor field.

[0159] When constructing a feature tensor field for multiple phenological stages based on orthogonal feature bases, the system fully preserves multi-dimensional correlation information such as time, space, and features, clearly presenting the relationships between various growth characteristics in different phenological stages. During tensor shrinkage of the feature tensor field, redundant dimensions are compressed, such as eliminating spatially repetitive dimensions unrelated to growth status, transforming the high-dimensional tensor field into a low-dimensional growth status vector. This vector concisely encapsulates the core growth characteristics of multiple phenological stages, facilitating subsequent integration with contribution weights and chronological arrangement to form a full life-cycle development trajectory. It also accurately conveys growth status information, providing an efficient data carrier for deducing physiological maturity paths and determining the optimal harvesting window.

[0160] S4. Based on the cumulative and specific contributions of the final harvest target in the *Hymenochloa chinensis*, establish a contribution quantification system for the *Hymenochloa chinensis* and assign contribution weights to the growth status vector.

[0161] In this embodiment of the invention, the step of establishing a contribution quantification system for the *Hymenochloa chinensis* based on the cumulative and specific contributions formed by the final harvest target, and assigning contribution weights to the growth status vector, includes:

[0162] The medicinal component content, biomass accumulation, and morphological development integrity of the *Hymenochloa chinensis* were extracted from historical harvesting data to form a quality characteristic set of the *Hymenochloa chinensis*.

[0163] The cumulative contribution of the multiple phenological stages to the set of quality characteristics is used as the first level of characterization, and the specific contribution of the multiple phenological stages to the key quality indicators in the set of quality characteristics is used as the second level of characterization to construct a multi-level contribution network for the Blood Leaf Orchid.

[0164] In the multi-level contribution network, the node centrality of the multi-phenological stage is evaluated, and the stage-based weights of the multi-phenological stage are determined.

[0165] Based on the similarity between the growth trend vector and the best harvested sample in the historical harvesting data, the basic weights of the stages are dynamically adjusted to obtain the contribution weights of the multi-phenological stages, and the contribution weights are assigned to the growth trend vector.

[0166] To extract the content of medicinal components, biomass accumulation, and morphological development integrity from historical harvesting data of *Hylocereus undatus*, and to form a quality characteristic set for *Hylocereus undatus*, harvesting records after each complete growth cycle of *Hylocereus undatus* were collected. From these records, information on the content of medicinal components directly related to medicinal value was screened out, including the specific accumulation of various effective active substances.

[0167] Simultaneously, biomass accumulation data reflecting plant growth were extracted from the records, covering intuitive growth indicators such as overall plant weight and volume. Furthermore, information on the morphological developmental integrity characterizing plant growth status was compiled, including appearance and structural features such as leaf quantity, stem thickness, and root development. These three types of information were categorized and organized by harvest batch, duplicate or invalid records were removed, and integrated to form a quality characteristic set that comprehensively reflects the harvested quality of *Hymenochloa chinensis*.

[0168] The cumulative contribution of multiple phenological stages to the concentration of quality characteristics is used as the first level of characterization, and the specific contribution of each phenological stage to the key quality indicators of the concentration of quality characteristics is used as the second level of characterization. When constructing the multi-level contribution network of *Cymbidium goeringii*, the cumulative effect of each phenological stage on various indicators of the concentration of quality characteristics is analyzed one by one. The cumulative contribution ratio of each stage to the content of medicinal components, biomass accumulation and morphological development integrity in the complete growth cycle is calculated, and these cumulative contribution information are used as the first level of characterization content.

[0169] Simultaneously, key quality indicators that play a decisive role in harvesting value are identified based on quality characteristics. The unique impact of each phenological stage on these key indicators is analyzed, such as the rapid accumulation of a certain core medicinal component or its key driving effect on biomass growth during a specific phenological stage. These specific contributions are used as the second-level characterization content. Using phenological stages as network nodes and hierarchical contributions as the basis for node association, a multi-level contribution network containing two characterization levels and reflecting the relationship between phenological stages and quality characteristics is constructed.

[0170] In a multi-level contribution network, when evaluating the node centrality of multiple phenological stages and determining the stage-based weights, the connection strength and influence range of each phenological stage node in the multi-level contribution network are analyzed to determine the degree of correlation between the node and other phenological stage nodes and quality characteristic indicators. The more quality characteristic indicators a node connects to and the greater its influence on other phenological stage nodes, the more prominent its core position in the network, and the higher its node centrality. The centrality value of each node is quantified using a fixed evaluation method, and corresponding stage-based weights are assigned according to the magnitude of the centrality value. The higher the centrality of a phenological stage, the greater its stage-based weight, ensuring that the basic weights accurately reflect the core role of the phenological stage in quality formation.

[0171] Based on the similarity between the growth status vector and the status of the best harvested sample in historical harvesting data, the basic weights of each stage are dynamically adjusted to obtain the contribution weights of each multi-phenological stage. When assigning contribution weights to the growth status vector, the best harvested sample with high medicinal component content, sufficient biomass, and complete morphological development is selected from historical harvesting data, and the corresponding growth status reference vector is extracted. The similarity between the current multi-phenological stage growth status vector and the reference vector is calculated, and the degree of fit between the two in terms of characteristic distribution and changing trends is analyzed.

[0172] If the similarity is high, it indicates that the growth state of the current phenological stage is close to the optimal harvesting standard, and the basic weight of this stage should be appropriately increased. If the similarity is low, the basic weight should be appropriately decreased according to the degree of difference. This dynamic adjustment method yields a contribution weight that more closely reflects the actual harvesting value. The adjusted contribution weights are then correlated one-to-one with each feature dimension of the growth trend vector, assigning a corresponding contribution weight to each feature dimension. This ensures that the growth trend vector accurately reflects the contribution value of different phenological stages to the final harvesting target.

[0173] The beneficial effects are that the content of medicinal components, biomass accumulation and morphological development integrity directly related to the harvesting value are extracted from historical harvesting data, and the resulting quality characteristic set accurately corresponds to the final harvesting target, providing a clear reference for contribution assessment and avoiding ineffective analysis that is divorced from actual harvesting needs.

[0174] A network is constructed using cumulative contribution and specific contribution as two-level representations. This reflects the long-term cumulative effect of multiple phenological stages on quality and highlights the unique impact of key stages on core quality indicators. It clearly presents the differentiated contributions of each stage and provides a scientific basis for weight allocation.

[0175] The node centrality of multiple phenological stages is evaluated based on a multi-level contribution network. The level of centrality directly corresponds to the coreness of the stage in quality formation, so that the basic weight of the stage can objectively reflect the importance of the stage and avoid the bias of subjective experience assignment.

[0176] By combining the growth status vector with the status similarity of the optimal harvested sample, the basic weights are adjusted so that the contribution weights can fit the actual growth status of the current plant. This ensures that the weights assigned to the growth status vector accurately match the stage contribution value, providing accurate data support for subsequent trajectory correction and maturity projection.

[0177] S5. Apply the contribution weight to correct the growth trend vector, and arrange the corrected results in chronological order, which is regarded as the full life cycle development trajectory of the Blood Leaf Orchid, so as to deduce the physiological maturity path of the Blood Leaf Orchid.

[0178] In this embodiment of the invention, the application of the contribution weight to correct the growth status vector, and the arrangement of the corrected results in chronological order, are regarded as the full life cycle development trajectory of the *Hymenochloa chinensis*, in order to deduce the physiological maturity path of the *Hymenochloa chinensis*, including:

[0179] The contribution weights are nonlinearly fused with the corresponding growth trend vectors to obtain the corrected growth trend vector of the *Hymenochloa chinensis*.

[0180] Arrange the corrected growth trend vectors in chronological order to obtain the growth trend sequence of the *Hypericum erythrophyllum*.

[0181] By eliminating the random fluctuation components of the growth status sequence, the full life cycle development trajectory of the *Hymenochloa chinensis* is obtained.

[0182] By performing trajectory curvature analysis on the entire life cycle development trajectory, the key turning points of trajectory changes in the entire life cycle development trajectory are obtained;

[0183] The key turning point region is mapped to the physiological maturity path of the Blood Leaf Orchid.

[0184] The process of eliminating the random fluctuation components of the growth pattern sequence to obtain the full life cycle development trajectory of the *Hymenochloa chinensis* includes:

[0185] Multi-scale intrinsic mode decomposition is performed on the growth trend sequence to obtain the multi-level frequency domain features of the growth trend sequence.

[0186] Extract the significant modes related to growth rhythm and the interference modes related to environmental noise from the multi-level frequency domain features;

[0187] The saliency mode and the interference mode are phase synchronized to obtain the reconstructed reference mode of the growth state sequence;

[0188] The random fluctuation components of the reconstructed baseline mode are eliminated, and the processed results are subjected to trajectory evolution to obtain the full life cycle development trajectory of the Bloodleaf Orchid.

[0189] When the contribution weights are nonlinearly fused with the corresponding growth status vectors to obtain the corrected growth status vector for *Cymbidium goeringii*, it is clear that the contribution weights have been allocated according to the contribution value of phenological stages to the harvest target. The growth status vector contains the core growth characteristics and trends of multiple phenological stages. Using a fixed nonlinear fusion method, the values ​​of each feature dimension in the growth status vector are adjusted based on the contribution weights. This strengthens the feature dimensions corresponding to phenological stages with high contribution and appropriately weakens the feature dimensions with low contribution. During the fusion process, the core structure of the growth status vector and the value orientation of the contribution weights are fully preserved, ensuring that the adjusted vector reflects both the actual growth state and highlights the contribution differences between different stages, ultimately forming the corrected growth status vector.

[0190] When arranging the corrected growth trend vectors in chronological order to obtain the growth trend sequence of *Cymbidium goeringii*, the phenological stage time nodes corresponding to each corrected growth trend vector are identified to ensure that each vector has a clear time identifier. Following the passage of time throughout the complete growth cycle of *Cymbidium goeringii*, all corrected growth trend vectors are arranged sequentially from front to back according to their time nodes, ensuring a natural connection between the end time of the phenological stage corresponding to the previous vector and the start time of the phenological stage corresponding to the next vector. During the arrangement process, the characteristic information of each vector remains intact, ultimately forming a growth trend sequence that can display the changes in growth trend over time.

[0191] To eliminate random fluctuations in the growth status sequence and obtain the full life cycle trajectory of *Illicium lanceolatum*, the characteristic values ​​of adjacent vectors in the growth status sequence are compared one by one to identify random fluctuations where values ​​suddenly jump and do not conform to the overall trend. These fluctuations are usually caused by non-essential factors such as short-term environmental disturbances. Using a fixed smoothing method, and referencing the characteristic values ​​before and after the fluctuations, the values ​​in the fluctuation areas are corrected to ensure that the corrected values ​​align with the overall growth trend while preserving the features in the sequence that reflect the essential changes in phenological stages. After smoothing, the growth status sequence shows a more stable and consistent trend, eliminating the influence of random disturbances and forming a full life cycle trajectory that truly reflects the growth and evolution of *Illicium lanceolatum*.

[0192] By performing trajectory curvature analysis on the entire life cycle development trajectory to identify key turning points in trajectory changes, the curvature of the trajectory is calculated segment by segment along the time axis of the entire life cycle development trajectory. The degree of trajectory curvature directly reflects the intensity of changes in growth status. When the curvature of the trajectory in a certain time period is significantly greater than that in adjacent time periods, it indicates that a drastic change in growth status has occurred during that time period, representing a significant turning point in growth. The trajectory regions corresponding to these time periods with significant curvature are marked; these regions are the key turning points in trajectory changes within the entire life cycle development trajectory, with each region corresponding to a significant adjustment in growth status.

[0193] When mapping key turning points to the physiological maturity path of *Illicium verum*, the changes in growth characteristics corresponding to each key turning point are analyzed to clarify the evolution of core indicators such as the content of medicinal components, biomass accumulation, and morphological development integrity within that region. Based on the changing patterns of these core indicators, the physiological maturity state of *Illicium verum* corresponding to the key turning points is determined. For example, the turning point from predominantly vegetative growth to predominantly reproductive growth corresponds to an important stage of physiological maturity enhancement. By connecting the physiological maturity states corresponding to all key turning points in chronological order, a path reflecting the evolution of the physiological state of *Illicium verum* from the early stage of growth to harvest maturity is formed. This path is the physiological maturity path of *Illicium verum*.

[0194] When performing multi-scale intrinsic mode decomposition (IMD) on a growth pattern sequence to obtain its multi-level frequency domain features, the sequence is decomposed layer by layer according to multiple preset scale levels, each scale level corresponding to a specific frequency range. The decomposition process follows the inherent variation law of the sequence, breaking it down into multiple independent intrinsic modes, each corresponding to a frequency domain feature component. Different scale levels focus on the short-term rapid changes and long-term slow changes of the sequence, respectively, ultimately yielding multi-level frequency domain features that cover different frequency ranges and comprehensively reflect the frequency domain characteristics of the sequence.

[0195] When extracting significant modes related to growth rhythm and interfering modes related to environmental noise from multi-level frequency domain features, it is first necessary to clarify that growth rhythm has periodic and stable characteristics, while environmental noise has random and irregular characteristics. Each intrinsic mode in the multi-level frequency domain features is analyzed one by one to determine whether its changes conform to the natural growth rhythm of *Hylocereus undatus*, such as whether it is consistent with the laws of phenological stage transitions and physiological index accumulation. Intrinsic modes that conform to these laws are considered significant modes. Intrinsic modes whose changes have no fixed period and are unrelated to growth rhythm are then screened out. These modes are caused by short-term environmental fluctuations and other interfering factors, and are thus considered interfering modes.

[0196] When synchronizing the phases of the saliency and interference modes to obtain the reconstructed reference mode of the growth trend sequence, the phase change pattern of the saliency mode is used as a benchmark to analyze the phase difference between the interference mode and the saliency mode. By adjusting the phase of the interference mode, the phase change of the interference mode is synchronized with that of the saliency mode, eliminating the feature conflict caused by phase misalignment. The phase-synchronized saliency and interference modes are then reintegrated according to their original frequency domain correlation to form a phase-coordinated and feature-unified mode, which serves as the reconstructed reference mode of the growth trend sequence.

[0197] The random fluctuation components of the reconstructed baseline mode were eliminated, and the processed results were subjected to trajectory evolution to obtain the full life cycle development trajectory of *Illicium lanceolatum*. Each data point in the reconstructed baseline mode was analyzed to identify random fluctuations that deviated from the overall trend of the mode and exhibited irregular fluctuations. A smoothing method was employed, using normal data before and after the fluctuations as a reference, to correct the fluctuations, ensuring that the corrected values ​​closely matched the overall evolution trend of the mode and completely eliminating the influence of random fluctuations. The corrected reconstructed baseline mode was then unfolded chronologically to simulate the evolution of *Illicium lanceolatum* during its growth process, forming a continuous, stable trajectory that truly reflects the core growth patterns. This trajectory represents the full life cycle development trajectory of *Illicium lanceolatum*.

[0198] The beneficial effect is that by nonlinearly fusing the contribution weight with the growth trend vector, the feature proportion of the high contribution stage is strengthened, and the corrected vector can accurately reflect the actual impact of each stage on growth, avoiding interference from irrelevant stage features with core information.

[0199] Arranging the corrected vectors in chronological order creates a growth pattern sequence that closely matches the temporal evolution of *Hylocereus undatus*, providing an ordered data foundation for constructing the full life cycle trajectory.

[0200] By eliminating random fluctuations in the growth trajectory sequence, the obtained full life cycle development trajectory can truly reflect the core evolutionary law of growth and avoid trajectory distortion caused by short-term environmental disturbances.

[0201] By analyzing the trajectory curvature, key turning points are identified and mapped to physiological maturity paths, enabling precise correlation between maturity changes and growth trajectory characteristics. This provides a clear basis for determining the harvest window based on maturity.

[0202] Multi-scale intrinsic mode decomposition yields multi-level frequency domain features, providing a clear frequency domain basis for distinguishing growth rhythms from environmental noise and avoiding confusion between the two types of information. Significant and interfering modes are extracted to clearly identify the core information reflecting growth rhythms and environmental noise interference, laying the foundation for trajectory optimization.

[0203] Phase synchronization of the two modes eliminates feature conflicts caused by phase misalignment, resulting in a reconstructed baseline mode that more accurately represents the growth trend. By removing random fluctuations and evolution trajectories from the reconstructed baseline mode, the obtained full life-cycle development trajectory truly reflects the core growth laws, providing high-quality data for subsequent maturity path projection.

[0204] S6. When the physiological maturity path reaches the historical best quality convergence interval of the Blood Leaf Orchid, it is determined as the best harvesting window for the Blood Leaf Orchid.

[0205] In this embodiment of the invention, determining the optimal harvest window for the *Hymenochloa chinensis* when the physiological maturity path reaches the historical best quality convergence interval includes:

[0206] At the optimal harvest period in the historical harvest data, the physiological maturity characteristics of the *Hymenochloa chinensis* are used as the quality characteristic benchmark for the *Hymenochloa chinensis*.

[0207] The matching degree between the physiological maturity path and the quality characteristic benchmark is monitored in real time. When the physiological maturity path enters the quality convergence critical region of the Blood Leaf Orchid, the stable state of the physiological maturity path in the historical best quality convergence range is confirmed.

[0208] Based on the stable dwell state, the convergence strength and duration of the physiological maturity path are comprehensively judged to obtain the optimal harvesting window for the Blood Leaf Orchid.

[0209] When considering the optimal harvest period from historical harvesting data, the physiological maturity characteristics of *Illicium lanceolatum* are used as the quality characteristic benchmark. First, samples with the highest levels of medicinal components, saturated biomass accumulation, and intact morphological development are selected from the accumulated historical harvesting data. The harvesting time corresponding to these samples is defined as the historical optimal harvest period. Next, the physiological maturity characteristics of these optimal harvest period samples are extracted, specifically including key indicators such as the accumulated concentration of core medicinal components, the actual accumulated biomass, the completeness of leaf and root development, and stem thickness. Then, the specific values ​​and performance status of these indicators are integrated into a unified standard system. This standard system serves as the quality characteristic benchmark for *Illicium lanceolatum*, providing a clear and comparable reference for subsequent matching monitoring.

[0210] Real-time monitoring of the matching degree between the physiological maturity path and the quality characteristic benchmark is conducted. When the physiological maturity path enters the quality convergence critical zone of *Illicium verum*, and it is confirmed that the physiological maturity path is in a stable state within the historically optimal quality convergence range, the quality convergence critical zone is first defined based on the fluctuation range of physiological maturity characteristics of samples from the historically optimal harvest period. The upper and lower limits of this critical zone correspond to the reasonable fluctuation boundaries of the physiological maturity characteristics of the optimal quality samples. Then, the current physiological maturity data of *Illicium verum* is collected at fixed time intervals. These real-time data are compared with each indicator in the quality characteristic benchmark one by one, and the degree of fit of each indicator is calculated and integrated into an overall matching degree value. When the matching degree values ​​collected for three consecutive times all fall within the quality convergence critical zone, it is determined that the physiological maturity path has entered the critical zone. Subsequently, the matching degree values ​​of the next five time intervals are monitored. If these values ​​remain within the critical zone and there is no obvious downward trend, it is confirmed that the physiological maturity path is in a stable state within the historically optimal quality convergence range.

[0211] Based on the stable residence state, a comprehensive judgment is made on the convergence strength and duration of the physiological maturity path to determine the optimal harvest window for *Illicium lanceolatum*. First, the convergence strength is evaluated by calculating the average deviation between the matching degree values ​​of each monitoring session and the quality characteristic benchmark under stable residence conditions. The smaller the average deviation, the higher the convergence strength. When the average deviation is less than a preset convergence strength threshold, the convergence strength is considered to be up to standard. Next, the convergence duration is evaluated by statistically analyzing the continuous residence time of the physiological maturity path within the historical best quality convergence interval. When the continuous residence time reaches a preset minimum duration threshold, the convergence duration is considered to be up to standard. When both convergence strength and convergence duration are up to standard, the time of achievement is taken as the start time of the optimal harvest window. Simultaneously, based on the average residence time of the best quality sample within the convergence interval in historical data, the end time of the optimal harvest window is determined. The time period consisting of the start time and the end time is the optimal harvest window for *Illicium lanceolatum*.

[0212] The beneficial effects are that by using the physiological maturity characteristics of the historically optimal harvest period as a quality benchmark, an objective and accurate reference standard is provided for subsequent matching degree monitoring, avoiding harvest judgments that deviate from the optimal quality target. Real-time monitoring of the matching degree between the maturity path and the benchmark confirms its stable residence within the optimal quality convergence range, ensuring that the growth state that meets the optimal quality is captured.

[0213] By combining the intensity and duration of convergence, we can ensure that the quality meets the standards at the time of harvest, and avoid harvesting too early or too late, thus effectively protecting the medicinal and economic value of *Hydrocotyle sibthorpioides*.

[0214] like Figure 2 The diagram shown is a functional block diagram of a system for predicting the harvest period of *Hymenochloa chinensis* that integrates multi-source data, provided by an embodiment of the present invention.

[0215] The *Gynostemma pentaphyllum* harvesting period prediction system 100, which integrates multi-source data according to the present invention, can be installed in an electronic device. Depending on the functions implemented, the *Gynostemma pentaphyllum* harvesting period prediction system 100 may include a multi-dimensional information field sensing module 101, a phenological stage division module 102, a growth status vectorization module 103, a contribution weight assignment module 104, a life cycle trajectory extrapolation module 105, and an optimal harvesting decision module 106. The modules described in this invention can also be referred to as units, which are a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.

[0216] In this embodiment, the functions of each module / unit are as follows:

[0217] The multidimensional information field sensing module 101 is used to synchronously sense and aggregate meteorological elements, soil profile physicochemical indicators and plant physiological signals of the blood-leaf orchid during its complete growth cycle, so as to obtain the spatiotemporal synchronous multidimensional information field of the blood-leaf orchid.

[0218] The phenological stage division module 102 is used to identify the critical point at which the growth pattern of the *Cymbidium goeringii* undergoes a fundamental change based on the evolution law of the spatiotemporal synchronous multidimensional information field, and to divide the complete growth cycle into multiple phenological stages of the *Cymbidium goeringii* based on the critical point.

[0219] The growth trend vectorization module 103 is used to perform dimensionality reduction processing on the multi-dimensional data of the multi-phenological stage, and to perform tensor synthesis on the trend features of the dimensionality reduction result to obtain the growth trend vector of the multi-phenological stage.

[0220] The contribution weight assignment module 104 is used to establish a contribution quantification system for the *Hymenochloa chinensis* based on the cumulative and specific contributions formed by the final harvest target in the *Hymenochloa chinensis*, and to assign contribution weights to the growth trend vector.

[0221] The life cycle trajectory extrapolation module 105 is used to apply the contribution weight to correct the growth trend vector and arrange the corrected results in chronological order as the full life cycle development trajectory of the Blood Leaf Orchid, so as to extrapolate the physiological maturity path of the Blood Leaf Orchid.

[0222] The optimal harvesting decision module 106 is used to determine the optimal harvesting window for the *Hymenochloa chinensis* when the physiological maturity path reaches the historical best quality convergence interval of the *Hymenochloa chinensis*.

[0223] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0224] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0225] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0226] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0227] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0228] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for predicting the harvesting period of Phalaenopsis by fusing multi-source data, characterized in that, The method comprises: S1, synchronously sensing and converging meteorological elements, soil profile physical and chemical indexes and plant body physiological signals of Aglaia odorata in a complete growth cycle to obtain a spatiotemporal synchronous multi-dimensional information field of the Aglaia odorata; S2, identifying a critical point of a fundamental change in a growth mode of the Aglaia odorata according to an evolution law of the spatiotemporal synchronous multi-dimensional information field, and dividing the complete growth cycle into multiple phenological stages of the Aglaia odorata according to the critical point; S3, performing dimension reduction processing on multi-dimensional data of the multiple phenological stages, and performing tensor synthesis on a situation feature of a result after the dimension reduction to obtain a growth situation vector of the multiple phenological stages, including: projecting high-dimensional features of the multi-dimensional data in the multiple phenological stages to a low-dimensional manifold space to obtain a feature manifold expression of the multiple phenological stages; decoupling local features in the feature manifold expression to obtain stable feature components of the feature manifold expression; recombining global features in the feature manifold expression to obtain dynamic feature components of the feature manifold expression; performing multi-scale feature cross fusion on the stable feature components and the dynamic feature components to obtain a multi-level feature expression of the feature manifold expression; eliminating redundant information in the multi-level feature expression to determine an orthogonal feature basis of the multi-level feature expression; constructing a feature tensor field of the multiple phenological stages according to the orthogonal feature basis, and performing tensor contraction on the feature tensor field to determine the growth situation vector of the multiple phenological stages; S4, establishing a contribution quantification system of the Aglaia odorata according to accumulative and specific contributions formed by a final harvesting target of the Aglaia odorata, and giving a contribution degree weight to the growth situation vector; S5, correcting the growth situation vector by using the contribution degree weight, and arranging the corrected result in time sequence, which is regarded as a whole life cycle development track of the Aglaia odorata, to deduce a physiological maturity path of the Aglaia odorata, including: performing nonlinear fusion on the contribution degree weight and a corresponding growth situation vector to obtain a corrected growth situation vector of the Aglaia odorata; arranging the corrected growth situation vector in time sequence to obtain a growth situation sequence of the Aglaia odorata; eliminating a random fluctuation component of the growth situation sequence to obtain the whole life cycle development track of the Aglaia odorata; performing trajectory curvature analysis on the whole life cycle development track to obtain a key turning area of trajectory change in the whole life cycle development track; mapping the key turning area to the physiological maturity path of the Aglaia odorata; S6, determining a best harvesting window of the Aglaia odorata when the physiological maturity path reaches a historical optimal quality convergence interval of the Aglaia odorata.

2. The method of claim 1, wherein the method is a method of fusing multi-source data for a leaf-cutting ant harvesting period prediction, characterized by, The method comprises: in a complete growth cycle of the Aglaia odorata, acquiring meteorological elements, soil profile physical and chemical indexes and plant body physiological signals of the Aglaia odorata; Eliminate the time inconsistency among the meteorological elements, the soil profile physical and chemical indexes and the plant body physiological signals, and obtain time standardized data of the Phreatia finetii; In the spatial dimension, register the geographical position information of the time standardized data, and obtain spatial alignment data of the Phreatia finetii; Integrate the attribute dimensions of different data sources in the spatial alignment data, and construct a spatiotemporal synchronous multi-dimensional information field of the Phreatia finetii.

3. The method of claim 1, wherein the method is a method of fusing multi-source data for a leaf-cutting ant harvesting period prediction, characterized by, According to the evolution law of the spatiotemporal synchronous multi-dimensional information field, identify the critical point of the fundamental change of the growth pattern in the Phreatia finetii, and divide the complete growth cycle into multiple phenological stages of the Phreatia finetii according to the critical point, including: Trend quantization is performed on the time dimension change among the meteorological elements, the soil profile physical and chemical indexes and the plant body physiological signals, and trend quantization factors of the Phreatia finetii are obtained; According to the trend quantization factors, the synergistic strength between different data sources in the spatiotemporal synchronous multi-dimensional information field is evaluated, and a trend synergy matrix of the Phreatia finetii is obtained; Characteristic evolution analysis is performed on the trend synergy matrix, and a dominant change mode of the trend synergy matrix is obtained; According to the stability change of the dominant change mode, the critical point of the fundamental change of the growth pattern in the Phreatia finetii is identified; According to the critical point, the complete growth cycle is divided into an initial phenological stage sequence of the Phreatia finetii; The phenological continuity of the initial phenological stage sequence is analyzed to adjust the boundary of the initial phenological stage sequence, and multiple phenological stages of the Phreatia finetii are obtained.

4. The method of claim 3, wherein the method further comprises: According to the trend quantization factors, the synergistic strength between different data sources in the spatiotemporal synchronous multi-dimensional information field is evaluated, and a trend synergy matrix of the Phreatia finetii is obtained, including: The trend quantization factors are decomposed into a meteorological element trend layer, a soil profile trend layer and a plant physiological trend layer; The first synergistic relationship strength between the meteorological element trend layer and the soil profile trend layer is analyzed to obtain a meteorological-soil synergistic relationship group of the Phreatia finetii; The second synergistic relationship strength between the soil profile trend layer and the plant physiological trend layer is evaluated to obtain a soil-physiological synergistic relationship group of the Phreatia finetii; The third synergistic relationship strength between the plant physiological trend layer and the meteorological element trend layer is converted by nonlinear mutual information to obtain a physiological-meteorological synergistic relationship group of the Phreatia finetii; The meteorological-soil synergistic relationship group, the soil-physiological synergistic relationship group and the physiological-meteorological synergistic relationship group are integrated in three dimensions to construct a trend synergy matrix of the Phreatia finetii, wherein the calculation formula of the trend synergy matrix is as follows: ； wherein is the trend synergy matrix, is a delay effect weight factor, is the weather-soil synergy group, is a tensor product operation, is the soil-physiology synergy group, is a matrix direct sum operation, is an immediate response weight factor, is the physiology-weather synergy group, is a non-linear adaptation weight factor.

5. The method of claim 1, wherein the method further comprises: According to the cumulative and specific contributions of the final harvesting target of the Phreatia finetii, a contribution quantization system of the Phreatia finetii is established, and a contribution degree weight is given to the growth tendency vector, including: The medicinal ingredient content, biomass accumulation degree and morphological development integrity of historical harvesting data of the Phreatia finetii are extracted to form a quality feature set of the Phreatia finetii; The cumulative contribution of the multiple phenological stages to the set of quality characteristics is represented as a first level, and the specific contribution of the multiple phenological stages to key quality indicators in the set of quality characteristics is represented as a second level, to construct a multi-level contribution network of the Cypripedium; In the multi-level contribution network, the node centrality of the multiple phenological stages is evaluated to determine the stage basic weight of the multiple phenological stages; According to the similarity of the growth trend vector and the trend of the optimal harvesting sample in the historical harvesting data, the stage basic weight is dynamically adjusted to obtain the contribution weight of the multiple phenological stages, and the contribution weight is given to the growth trend vector.

6. The method of claim 1, wherein the method further comprises: The elimination of the random fluctuation component of the growth trend sequence to obtain the full life cycle development trajectory of the Cypripedium includes: Multi-scale intrinsic mode decomposition is performed on the growth trend sequence to obtain multi-level frequency domain features of the growth trend sequence; The significant mode related to the growth rhythm and the interference mode related to the environmental noise in the multi-level frequency domain features are extracted; The significant mode and the interference mode are subjected to phase synchronization processing to obtain a reconstructed reference mode of the growth trend sequence; The random fluctuation component of the reconstructed reference mode is eliminated, and the trajectory evolution is performed on the processed result to obtain the full life cycle development trajectory of the Cypripedium.

7. The method of claim 5, wherein the method further comprises: When the physiological maturity path reaches the historical optimal quality convergence interval of the Cypripedium, the optimal harvesting window of the Cypripedium is determined, which includes: In the optimal harvesting period in the historical harvesting data, the physiological maturity characteristics of the Cypripedium are taken as the quality characteristic reference of the Cypripedium; The matching degree change of the physiological maturity path and the quality characteristic reference is monitored in real time, and when the physiological maturity path enters the quality convergence critical zone of the Cypripedium, the stable residence state of the physiological maturity path in the historical optimal quality convergence interval is confirmed; Based on the stable residence state, the convergence intensity and convergence duration of the physiological maturity path are comprehensively judged to obtain the optimal harvesting window of the Cypripedium.

8. A system for predicting the harvesting period of Phalaenopsis by fusing multi-source data, characterized in that, A system for implementing the Cypripedium harvesting period prediction method of fusing multi-source data according to claim 1, the system comprising: A multi-dimensional information field perception module is used to synchronously perceive and converge meteorological elements, soil profile physicochemical indicators, and plant physiological signals of Cypripedium in the complete growth cycle to obtain a spatio-temporal synchronous multi-dimensional information field of the Cypripedium; A phenological stage division module is used to identify critical points of fundamental changes in the growth mode of the Cypripedium according to the evolution law of the spatio-temporal synchronous multi-dimensional information field, and divide the complete growth cycle into multiple phenological stages of the Cypripedium according to the critical points; A growth trend vectorization module is used to perform dimension reduction processing on the multi-dimensional data of the multiple phenological stages, and perform tensor synthesis on the trend characteristics of the dimension-reduced result to obtain a growth trend vector of the multiple phenological stages; A contribution weight assignment module is used to establish a contribution quantification system of the Cypripedium according to the cumulative and specific contribution of the final harvesting target of the Cypripedium, and assign a contribution weight to the growth trend vector. a life cycle trajectory deduction module configured to correct the growth trend vector by applying the contribution weight, and arrange the corrected result in time sequence, and regard the result as a full life cycle development trajectory of the Phreatia finetii, so as to deduce a physiological maturity path of the Phreatia finetii; an optimal harvesting decision module configured to determine a best harvesting window of the Phreatia finetii when the physiological maturity path reaches a historical optimal quality convergence interval of the Phreatia finetii.

Images